Approach, avoidance, and context
Does context influence the link between affect and approach and avoidance
tendencies?
Myrthe Knetemann
Student number: 10645780
Date: 24 March, 2014 – 31 July, 2015
Number of EC’s: 26
Study: Research Master Brain and Cognitive Sciences University of Amsterdam
Track: Cognitive Science
Supervisor: Dr. R. H. Phaf
Brain and Cognition
Psychology Research Institute University of Amsterdam Co-assessor & UvA Representative: Dr. M. Rotteveel
Social Psychology
Psychology Research Institute University of Amsterdam
Abstract
The aim of this study was to investigate the influence of context on the link between affect and approach and avoidance tendencies. Initial research with the approach-avoidance task suggested a direct, automatic, link (Chen & Bargh, 1999) between affect and arm flexion and extension. Rotteveel and Phaf (2004), however, concluded to an indirect link because no effects were found with implicit evaluation. In addition, opposite directions for the effect emerged with self and object-referenced interpretations (Phaf, Mohr, Rotteveel and Wicherts, 2014). Context has been known to play an important role in the link between affect and action tendencies (Paulus & Wentura, 2013). It was investigated here whether context could modulate the link between affect and arm flexion and extension. Pictures of facial expressions were explicitly evaluated with the flexion-extension button stand, which measures approach and avoidance tendencies. Ambiguous, affectively neutral, context pictures provided different contexts for the interpretation of the facial expressions. A pseudo-candid photo manipulation (Mobbs et al., 2006) was used to strengthen the impression that the facial expressions reacted to the preceding context pictures. The different contexts revealed only, statistically unreliable, modulation of the approach and avoidance tendencies towards the happy and angry faces. Further research with improved setups could well yield stronger evidence for a role of contextual interpretation in the link between affect and action tendencies.
Introduction
Initial research with the approach-avoidance task (AAT) suggested a direct, automatic link between affect and approach and avoidance tendencies, even with flexion and extension arm movements (Chen & Bargh, 1999; Duckworth, Bargh, Garcia, & Chaiken, 2002; Solarz, 1960). Emotions are considered responsible for preparing for swift and direct action without explicit deliberation (Lang, Bradley, & Cuthbert, 1990). Rotteveel and Phaf (2004), however, argued that automatic affective evaluation does not automatically predispose for approach and avoidance action tendencies. Their results indicated that when affective stimuli were explicitly evaluated (e.g., by judging something as positive or negative) the results were the same as those of Chen and Bargh, but when affective stimuli were evaluated implicitly (e.g., by judging the gender of the emotional expressions) no effect on approach and avoidance action tendencies was obtained. A meta-analysis by Phaf, Mohr, Rotteveel and Wicherts (2014) also supports an indirect link between affect and action tendencies. They included 29 studies investigating action tendencies towards affective stimuli and concluded to an indirect link because no evidence was found for an effect with implicit evaluation and opposite directions for the effect emerged with self and object-referenced interpretations.
Context has been shown to play an important role as a moderator in the link between affect and action, as for instance in a study by Paulus and Wentura (2013). They investigated the influence of group membership of the expresser on approach and avoidance tendencies to facial expressions. They manipulated group membership by showing German participants pictures of White-Caucasian and Middle-Eastern young men (in-group and out-group, respectively). Paulus and Wentura found that facial expressions (fear and happiness) expressed by in-group members elicited congruent action tendencies (avoidance and approach respectively), whereas facial expressions expressed by out-group members elicited the opposite pattern. Paulus and Wentura demonstrated that a neutral context (ethnicity) could
influence the interpretation of an affective facial expression. However, they used facial expressions of different persons to show opposing action tendencies to affect. The current study will try to elicit different action tendencies to the facial expression of one and the same person.
The aim of this study is to investigate whether the link between affect and approach and avoidance tendencies is based on an interpretation of the context in which the affective stimuli appear (i.e., indirect). This leads to the following research question: do contextual interpretations influence the link between affect and approach and avoidance tendencies?
This research focuses on a manipulation of the link between affect and action tendencies, while assuming that positive and negative affect is elicited by happy and angry facial expressions (Rotteveel & Phaf, 2004), and that approach and avoidance tendencies can be measured through explicit evaluation with the vertical button stand (Rotteveel & Phaf, 2004; Phaf, Mohr, Rotteveel, & Wicherts, 2014). Context is manipulated by showing ambiguous, affectively neutral, pictures, which are disambiguated by the facial expressions. To strengthen the impression that the emotional expressions react to the preceding context pictures, a pseudo-candid photo manipulation is used (Mobbs et al., 2006) based on the Kuleshov Effect. Participants are told that the pictures of the facial expressions were made earlier in another study. In this study they showed people pictures and subsequently asked them to express their feelings about these pictures, while their facial expressions were recorded. A pseudo-candid photo is used to illustrate this process (see Figure 1). Participants were asked to evaluate the recorded facial expressions explicitly as positive or negative. Behavioral studies have shown that the effect of context is higher when the clarity of the facial expression is lower (Ekman et al., 1982; Trope, 1986). Therefore we will use morphed faces displaying subtly happy and angry facial expressions.
Figure 1. The pseudo-candid photo manipulation was strengthened by showing a picture of an actor viewing a context picture on a monitor, and a webcam recording expressive facial responses of the person to this picture.
Our hypothesis is that the same facial expressions automatically eliciting specific affect either facilitate approach or avoidance tendencies due to contextual interpretation. This means that happy facial expressions (corresponding to positive affect) will foster approach tendencies in affiliative contexts but avoidance tendencies in threatening contexts. And angry facial expressions (corresponding to negative affect) raise approach tendencies in adversarial contexts but avoidance tendencies in threatening contexts. More specifically, we hypothesize that happy facial expressions will foster approach tendencies in the ‘people walking towards’ and the ‘food’ context and avoidance tendencies in the ‘people walking away’ and the ‘police’ context. And angry facial expressions will raise approach tendencies in the ‘people walking towards’, ‘people walking away’ and ‘police’ context and avoidance tendencies in the ‘food’ context. Approach tendencies will be measurable by shorter flexion reaction times with positive than with negative evaluations, and avoidance tendencies by shorter extension reaction times with negative than with positive evaluations. It is expected that the influence of affect on latency times will be larger on initiation times than on movement times (Rotteveel & Phaf, 2004).
Method
Participants
Forty-five Dutch speaking students participated in a one-hour session for course credits (14 participants), financial compensation (9 participants) or voluntarily (22 participants). The data of four participants were excluded, one for not following instructions and the others for having more than 20% incorrect responses on the relatively easy face evaluation task. The remaining 41 participants (27 female) all had normal or corrected-to-normal vision, were right-handed, and signed informed consent. The age range was between 18 and 26 years with a mean of 22.3 years (SD = 2.03). The experiment was announced as “Blije situaties?” (“Happy situations?”), and in the general introduction it was told that we were investigating the relationship between emotion and action.
Design
The experiment had a 4 (Context: food vs. people towards vs. people walking away vs. police) x 2 (Facial expressions: happy vs. angry) x 2 (Response instruction: congruent vs. incongruent) within-subjects factorial design. The experiment had 64 experimental trials (four replications each of the 16 conditions) and 16 practice trials (8 trials before each of the two different response instruction blocks). Because each trial required a unique context picture and the facial expression of a unique model, 20 different pictures for each context, and 40 different pictures for each facial expression were used. Each context was randomly coupled to a facial expression. Context and facial expressions were randomized within blocks (including the practice blocks) and response instruction was counterbalanced between blocks. Average initiation times (releasing the home button) and movement times (pushing the response buttons) over correct responses were calculated with explicit evaluation with the vertical button stand (Rotteveel & Phaf, 2004). Initiation times (ITs) that were lower than 300 ms or
higher than 1000 ms and movement times (MTs) that were higher than 400 ms were excluded from the analysis. Trials where participants released the home button too fast (during the presentation of the context) started over again and were included in the analysis.
Material
The context stimuli consisted of ambiguous, affectively neutral, pictures in four different categories: food, people walking towards, people walking away, and police (see Figure 2 and Appendix A). All pictures were derived from the Internet through Google's search engine and were selected to meet predefined criteria. In the 'food' category a plate with food and cutlery was shown from above or diagonally above. The food on the plate was blurred with the Gaussian blur filter in Photoshop CS4 (see Figure 2). This blurring was done to minimize affective reactions to the food itself. In the 'people walking towards' category one or more people were shown walking towards the viewer. The persons in these pictures should show an as neutral as possible emotion on their faces. The 'people walking away' category contained pictures of one or more people walking away from the viewer. And the 'police' category contained pictures of one or more police officers talking to one or more persons, in this category it was also aimed that persons showed an as neutral as possible emotion on their faces.
Figure 2. Examples of the pictures that were used in the different context conditions.
The face stimuli consisted of happy and angry faces (with a closed mouth) derived from the NIM-STIM Set of Facial Expressions (Tottenham et al., 2009). The automatic "Face Locator" of FantaMorph 5 (Deluxe Edition) was used to generate morphs between neutral and happy or neutral and angry facial expressions (see Figure 3 and Appendix B). This resulted in two more subtle facial emotions (i.e., 50% happy and 50% angry). In both conditions 18 female models and 22 male models were used.
Figure 3. An example of a neutral facial expression and two computerized morphs.
Apparatus
The stimuli were presented on a 22-inch, 1680 by 1050 pixel resolution monitor (ASUS VW222U). The stimuli were scaled to a width of 800 pixels (26.6 cm) while maintaining the aspect ratio. Stimuli were presented using Presentation®
software (Version 17.0, www.neurobs.com). Participants were seated in front of the monitor with a distance of one meter between the eyes and the monitor. The horizontal visual angle of the stimuli was 15.2 degrees. The light in the experimental room was dimmed.
Responses could be given with two one-button boxes that were fixed to a vertical three-button stand (Rotteveel & Phaf, 2004, see Figure 4). Participants were seated left of the stand (on a chair with a height of 45 cm) and had to operate the stand with their right hand. They first had to press (and hold) the home button loosely with the back of their right hand (starting position). The height of the home button was adjusted for each participant individually, so that the angle between the lower arm and the upper arm was 110° for all participants. In this
way, both biceps en triceps are equally tensed (i.e., in an isometric position) when holding the home button. Subsequently participants were instructed to push the upper or lower button in response to a stimulus. The two response buttons were positioned perpendicularly above and below the home button (with a distance of 10.3 cm), in this way participants could simply flex or extend their arm without the need to precisely aim at the response buttons. They had to push the response buttons with the top or bottom side of their hand (not turning their hand when responding). Pressing the upper button and the lower button corresponded to arm flexion and extension, respectively.
Figure 4. The three-button vertical stand, the two response buttons were positioned perpendicularly above and below the home button (fixed in the middle).
Each trial started when participants pushed the home button and held it pressed. After a jittered time interval (1-3 s) a fixation-cross appeared in the middle of the screen (500 ms). Subsequently the context picture was shown (4 s), and followed immediately by a facial expression. The target face disappeared on response.
Procedure
Participants were told that in each trial they would see a picture followed by a photo of a facial expression, and that there was a connection between the two. During the pseudo-candid photo manipulation participants were explained that each facial expression expressed how
that person felt about the preceding picture. Participants were instructed to evaluate (i.e., positively or negatively) the facial expression of the person as fast and accurately as possible with the vertical button stand. It was clearly mentioned that they needed to evaluate the facial expression, and not the facial expression in relation to the preceding context.
In the experimental blocks they received either an affect-congruent instruction (pressing the upper button with positively valenced faces and the lower button with negatively valenced faces) or an affect-incongruent instruction (pressing the upper button with negatively valenced faces and the lower button with positively valenced faces). Twenty-one participants started with a block where they had to give congruent responses and subsequently proceeded with a block where they had to give incongruent responses. The other participants followed the reversed order. The two experimental blocks were both preceded by practice block (8 trials) to familiarize participants with the response instruction. Also, participants performed an additional practice block of 10 trails at the beginning of the experiment to get familiar with the responses of the vertical button stand. In this extra practice block a picture of a neutral landscape was shown, followed immediately by an arrow pointing upwards or downwards. Participants had to press the upper or lower button (in response to an upwards or downwards arrow, respectively) as fast and accurately as possible.
Between the two experimental blocks, a digital game of memory was completed to ease the transition between response instructions. In the memory game 30 cards (of which 15 couples had an identical picture on it, see appendix C) were shown with their faces down, and participants could turn two cards at a time to check if they were the same. If they were the same, the cards would stay lying with their faces up, and if they were wrong the cards turned upside down again. After this participants could turn another couple of cards. This continued until all the couples were found or the maximum period of 4 minutes was reached.
Participants performed all the trials with the experimenter present, to ensure that they followed the instructions. The experiment was concluded with an exit interview in which participants were asked how they experienced the experiment and what strategies they had used.
Results
Participants reported that they were overall able to correctly evaluate the facial expressions as positive or negative. A few participants reported that not all faces looked natural to them, due to the morphing. Many participants reported that they thought that the faces often responded a bit strange to the preceding context. This confirms that our pseudo candid photo manipulation worked, people were interpreting the faces in relation to the preceding context picture. Also it was reported often that the incongruent response instruction (giving a positive evaluation with arm extension and a negative evaluation with arm flexion) felt more difficult and less natural.
Hundred ninety-seven (8%) trials were excluded from the analysis because participants gave an incorrect evaluation. Seventy-seven initiation times (3.1%) were excluded from the analysis because they were shorter than 300 ms or longer than 1000 ms. Seventy-nine movement times (3.2%) were excluded from the analysis because they were higher than 400 ms. The results were analyzed using a 4 (Context: food vs. people walking towards vs. people walking away vs. police) by 2 (Face: positive vs. negative) repeated measures ANOVA on the flexion ITs, extension ITs, flexion MTs and extension MTs.
Initiation times (ITs)
Table 1. Mean (SD) reaction times for home-button release (IT).
Flexion Extension
Food Positive 639 (55) 671 (84)
Negative 659 (97) 639 (76)
People towards Positive 652 (78) 684 (82)
Negative 678 (98) 653 (82)
People away Positive 644 (74) 662 (88)
Negative 679 (98) 637 (84)
Police Positive 674 (86) 683 (97)
Despite the morphing of the emotional faces to more subtle expressions, clear congruent priming effects showed up in the preparation of approach-avoidance movements (cf. Rotteveel & Phaf, 2004). Participants were overall faster to release the home button to perform a positive evaluation with arm flexion than a negative evaluation with arm flexion (see Table 1 and Figure 5). Participants were also overall faster to release the home button to perform a negative evaluation with arm extension than a positive evaluation with arm extension (see Table 1 and Figure 6). The main effect of face valence was marginally significant in the flexion ITs (F(1, 40) = 2.909, p = 0.096, = 0.068) and fully significant in the extension ITs (F(1, 40) = 10.468, p = 0.002, = 0.207). The effect size for the flexion ITs can be called small and for the extension ITs medium.
Figure 5. Means and standard errors for initiation times (ITs) with arm flexion.
ηp 2 ηp 2 620 630 640 650 660 670 680 690 700
Food People Towards People Away Police
Pos Neg
Figure 6. Means and standard errors for initiation times (ITs) with arm extension.
Table 2. Mean (SD) overall reaction times per context for home-button release (IT).
Flexion Extension
Food 649 (63) 655 (65)
People towards 665 (74) 668 (70)
People away 661 (70) 650 (72)
Police 674 (80) 662 (75)
The different contexts also showed appreciable differences in flexion and extension ITs, aggregated over happy and angry faces (see Table 2). The fastest preparation for flexion, irrespective of face valence, was made with food contexts, followed by the people walking away, the people walking towards and the police contexts (see Figure 7). The fastest overall extension ITs emerged for the people walking away contexts, followed by the food, the police and the people walking towards contexts (see Figure 8). The main effect of context on flexion ITs was significant (F(3, 120) = 3.268, p = 0.024, = 0.076). The effect size was however small. Post-hoc tests using the Bonferroni correction revealed that the differences between the food contexts and the police contexts were fully significant (p = 0.025). The main effect of context was not significant for the extension ITs (F(3, 120) = 1.365, p = 0.257, = 0.033). This effect size was also small.
620 630 640 650 660 670 680 690 700
Food People Towards People Away Police
Pos Neg ηp 2 ηp 2
Figure 7. Means and standard errors per context for initiation times (ITs) with arm flexion.
Figure 8. Means and standard errors per context for initiation times (ITs) with arm extension.
There was no significant interaction between context and face for the flexion ITs (see Figure 5, F(3, 120) = 1.553, p = 0.204, = 0.037), and also not for the extension ITs (see Figure 6, F(3, 120) = 0.486, p = 0.693, = 0.012). In other words, the main effect of face did not change reliably for the different levels of context. But despite the absence of a reliable interaction, interesting results can be seen in the data.
620 630 640 650 660 670 680 690 700
Food People Towards People Away Police
620 630 640 650 660 670 680 690 700
Food People Towards People Away Police
ηp 2
ηp 2
Table 3. Mean difference of reaction times and effect size for home-button release (IT) between positive flexion and negative flexion and positive extension and negative extension per context.
Flexion Cohen’s d Extension Cohen’s d
Food 19 0.25 32 0.40
People towards 26 0.29 31 0.38
People away 35 0.41 25 0.29
Police 1 0 44 0.50
Despite the absence of a reliable interaction, in absolute value interesting differences in priming can be seen between different contexts. Modulation by the faces appeared to be stronger for the extension ITs than for the flexion ITs (see Table 3 and Figure 5 and 6), but the variation in this modulation was larger with flexion than with extension. With flexion in the police context the difference between happy and angry faces seemed to vanish (d = 0). With extension, the police context yielded the largest priming of faces of all contexts (d = 0.50). Apparently, the police context biased so overwhelmingly to avoidance and extension that the presentation of happy or angry faces could not modulate the slow execution with arm flexion any further. The opposite pattern seemed to occur with the people walking away context, although to a lesser extent. For this context the difference between positive and negative faces was the largest of all contexts in the flexion ITs (d = 0.41), but the smallest in the extension ITs (d = 0.29). Apparently the people moving walking context applied more strongly to approach and flexion than to avoidance and extension preparation. The food and the people walking towards contexts seemed to apply more strongly to extension than to flexion preparation, although these effects were less strong. Finally, although a happy face clearly facilitated approach to food, the angry face did not slow down approach to the same extent as for the other contexts in the flexion ITs.
Movement Times (MTs)
Table 4. Mean (SD) movement times (MT).
Flexion Extension
Food Positive 152 (45) 163 (55)
Negative 160 (62) 149 (36)
People towards Positive 143 (45) 152 (43)
Negative 153 (53) 152 (38)
People away Positive 145 (55) 162 (48)
Negative 152 (56) 152 (35)
Police Positive 153 (47) 162 (55)
Negative 160 (59) 148 (41)
In the movement times (MTs) there was also a congruent priming effect in the execution of approach-avoidance movements. Participants were overall faster to perform a positive evaluation with arm flexion than a negative evaluation with arm flexion (see Table 4 and Figure 9). Participants were also faster to perform a negative evaluation with arm extension than a positive evaluation with arm extension (see Table 4 and Figure 10). The main effect of face was marginally significant in the flexion MTs (F(1, 40) = 2.974, p = 0.092, = 0.069) and also in the extension MTs (F(1, 40) = 3.945, p = 0.054, = 0.09). The effect sizes were small for both flexion and extension MTs. The extension effect sizes were larger than the flexion effect sizes in both initiation and movement times, although a substantially larger effect size was found for the extension ITs.
ηp
2 η
p 2
Figure 9. Means and standard errors for movement times (MTs) with arm flexion.
Figure 10. Means and standard errors for movement times (MTs) with arm extension.
Table 5. Mean (SD) overall movement times (MT) per context.
Flexion Extension Food 156 (48) 156 (39) People towards 148 (46) 152 (34) People away 148 (52) 157 (36) Police 157 (48) 155 (42) 140 145 150 155 160 165 170
Food People Towards People Away Police
Pos Neg 140 145 150 155 160 165 170 175
Food People Towards People Away Police
Pos Neg
The effect of the different contexts, irrespective of face, again appeared to be larger in the flexion MTs than in the extension MTs (see Table 5). The fastest executions of flexion were made with the people walking towards and the people walking away contexts, followed by the food and police contexts (see Figure 11). The fastest overall extension MTs emerged for the people walking towards contexts, followed by the police, the food and the people walking away contexts (see Table 5 and Figure 12). The main effect of context on flexion MTs was significant (F(3, 120) = 3.649, p = 0.015, = 0.084). The effect size can however be called small. Post-hoc tests using the Bonferroni correction revealed that the differences between these contexts were significant between the food contexts and the people walking towards contexts (p = 0.046). The main effect of context was not significant for the extension MTs (F(3, 120) = 0.482, p = 0.696, = 0.012). This effect size can also be called small.
Figure 11. Means and standard errors per context for movement times (MTs) with arm flexion.
ηp 2 ηp 2 140 145 150 155 160 165 170 175
Figure 12. Means and standard errors per context for movement times (MTs) with arm extension.
Similar to the initiation times, there was no significant interaction between context and face for the flexion MTs (F(3, 106.549) = 0.061, p = 0.971, = 0.002) and also not for the extension MTs (F(3, 120) = 1.003, p = 0.394, = 0.024). The main effect of face did not change reliably for the different contexts, but again, interesting results did seem to occur in the data.
Table 6. Mean difference and effect size for movement times (MT) between positive flexion and negative flexion and positive extension and negative extension per context.
Flexion Cohen’s d Extension Cohen’s d
Food 7 0.13 14 0.29
People towards 10 0.20 1 0.02
People away 7 0.12 10 0.24
Police 7 0.14 14 0.28
Differential priming effects of the contexts on the faces were also present in the movement times. The priming effects again appeared to be stronger for the extension MTs than for the flexion MTs (see Table 6 and Figure 9 and 10). Furthermore, the strongest priming effect with flexion (d = 0.20) corresponded with the weakest priming effect with
140 145 150 155 160 165 170 175
Food People Towards People Away Police
ηp 2
ηp 2
extension (d = 0.02) for the people towards context. The people towards context biased so overwhelmingly to approach and flexion that the presentation of happy or angry faces could not modulate the slow extension execution any further. Although to a lesser extent, the opposite pattern seemed to occur with the other three contexts. The interpretation of happy and angry faces in these contexts apparently applied more strongly to avoidance and extension than to approach and flexion execution.
Discussion
Independent of context, initiation times and movement times were faster with affect-congruent responses than with affect-inaffect-congruent responses to the target faces. People were faster to perform a positive evaluation with arm flexion than a negative evaluation with arm flexion, and they were faster to perform a negative evaluation with arm extension than a positive evaluation with arm extension. These results replicate the explicit evaluation results of Rotteveel & Phaf (2004, Experiment 1) with a slightly different experimental set-up. There were some small differences between our results and the results of Rotteveel and Phaf. Our effect of face was larger for extension than for flexion reaction times, whereas Rotteveel and Phaf found a larger effect with flexion than extension. We also obtained similar effects in movement times and initiation times, whereas Rotteveel and Phaf found large effects in initiation times but no effects in movement times.
Irrespective of face valence, there were also reliable differences found in approach and avoidance tendencies to the different contexts. The food context in particular was more biased to arm flexion and approach than the other contexts in the initiation times. And the people walking towards and the people walking away contexts were more biased to arm flexion and approach than the food and police context in the movement times. The different contexts were selected to be as neutral as possible, so there is no particular reason to think that these contexts directly caused different approach and avoidance tendencies. Therefore it is more likely that these differences support our hypothesis that the contexts were used for the interpretation of the facial expressions.
There were substantial differences visible in difference times between positive flexion and negative flexion and positive extension and negative extension per context. The police context (in the initiation times) was biased so overwhelmingly to extension and avoidance, that positive and negative faces could not modulate the slow flexion execution any further.
And the people walking towards context (in the movement times) was so biased to flexion and approach, that positive and negative faces could not modulate the slow extension execution anymore. These results suggest that, although participants overall responded faster with affect congruent responses to the faces, context also appeared to be involved in the interpretation of the faces. So although we did not find that the context pictures redirected the relationship between affect and approach and avoidance tendencies, the results do support a substantial influence of context on the link between affect and approach and avoidance tendencies.
The aim of this study was to investigate whether the link between affect and approach and avoidance tendencies is based on an interpretation of the context in which the affective stimuli appear. Finding an influence of context would support an indirect link between affect and approach and avoidance, instead of a direct, automatic link (Chen & Bargh, 1999). Our results do support an influence of context: irrespective of face, reliable differences were found in approach and avoidance tendencies to different contexts, and the difference scores between positive and negative faces also substantially differed between different contexts. These effects were however small. With the power expected from these small effects, statistical reliability cannot reasonably be expected. The conclusion to the null-hypothesis of no difference, however, would also not be warranted (Dienes, 2014). So we will further consider the results in the light of the psychological hypotheses instead of the statistical hypotheses. The alternative psychological hypotheses are a direct or indirect link between affect and approach-avoidance tendencies. A direct link would mean that positive and negative affect automatically lead to approach and avoidance tendencies, respectively. In our study, faces indeed congruently modulated approach and avoidance tendencies. But this does not imply a direct link between affect and approach and avoidance, because the participants explicitly evaluated the faces with the button stand (cf. Phaf et al., 2014). The meta-analysis of Phaf et
al. (2014) showed that no significant effect sizes were found for implicit evaluation with the approach-avoidance task, whereas medium effect sizes were found for explicit evaluation, irrespective of self or object-referenced evaluation. These results indicate that positive and negative affect do not automatically cause approach and avoidance tendencies, respectively, but that the interpretation of the movements is important. Eder (2011) found that the approach-avoidance task does not directly measure action tendencies, but implementation intentions: “Temporary links between affective cues and goal-directed responses that are set up as a result of task instructions.” If no congruency effects are found with implicit evaluation, and if congruent implementation intentions are needed to do find congruency effects with explicit evaluation, this supports an indirect link between affect and approach tendencies. An indirect link would further mean that context could have an influence on the link between affect and approach and avoidance, and our results do show that context has an influence on the interpretation of the positive and negative faces in terms of approach and avoidance. We therefore conclude to an indirect link between affect and approach and avoidance tendencies, and suggest some improvements to the experimental design so that on the one hand the effects will become larger and on the other hand even the small effects may become statistically reliable (i.e., the measurements become more accurate).
First of all, there was a lot of noise in the data of this study, which was probably mainly caused by the small number of trials for each participant. It was expected that the approach-avoidance task with the vertical button stand would be very exhausting for participants, because they needed to hold their arms without support for prolonged periods of time. This led us to the decision to have a smaller number of trails per participant. But in the future, it would be better to have more trials in each condition. If there are enough breaks during the experiment, participants are well able to perform these extra trials. It would also be an option to compare less context conditions to each other; this will lead to a larger number of trials for
each context condition. Second, in this study we started directly with testing the approach and avoidance tendencies to facial expressions in combination with a context picture. While in the future it would be better to these those conditions separately first. In this way, the effect of context can be measured as the difference between action tendencies to facial expressions with and without a context. This would be a more “pure” measure of the influence of context. Also, in this way it is made possible to look at the link between facial expressions and action tendencies and context and action tendencies separately. This is interesting because in the current study we could not differentiate if the measured action tendencies were mainly caused by the facial expressions or by the interpretation of the context. Finally, in this study the most notable priming effects were seen with the food context, the police context and the people walking towards contexts. The aim of the current study was to look at the effect of a neutral context on the link between affect and action tendencies, as Paulus and Wentura (2013) did in their study with ethnicity. However, if we take a better look at the pictures from the context conditions, it can be argued that the context conditions that showed the largest priming effects also contained more affective information. The people walking towards and police contexts, for example, contained facial expressions of people. While it is known that facial expressions have a very strong relationship with affect. We did try to control for emotional facial expressions in het context pictures, but we did not test the approach and avoidance tendencies to context pictures separately, so we cannot be sure about this. With food, on the other side, people also may experience a very strong relation with affect. This leads us to the hypothesis that maybe more affective priming is necessary to redirect the link between affect and approach and avoidance tendencies. Maybe the “neutral” manipulation as used by Paulus and Wentura was not so neutral after all, and thus “prejudices” are stronger affectively loaded than we think they are. In a next study, it is therefore advised to use context pictures with a higher affective value, so maybe larger effect sizes could be found.
The general intuition was that positive and negative affect automatically cause approach and avoidance tendencies, respectively. This study however shows that context does seem to have an influence on the link between approach and avoidance. Apparantly, more conscious (or nonconscious) interpretation of the situation takes place, before taking action. Even a neutral characteristic such as ethnicity, can cause that some people act differently to affectively similar facial expressions.
Appendix A
Appendix B
Appendix C
References
Chen, M., & Bargh, J. A. (1999). Consequences of automatic evaluation: Immediate
behavioural predispositions to approach or avoid the stimulus. Personality and Social
Psychology Bulletin, 25, 215-224.
Dienes, Z. (2014). Using Bayes to get the most out of non-significant results. Frontiers in
psychology, 5, 781.
Duckworth, K. L., Bargh, J. A., Garcia, M., & Chaiken, S. (2002). The automatic evaluation of novel stimuli. Psychological Science, 13, 513-519.
Eder, A. B. (2011). Control of impulsive emotional behavior through implementation intentions. Cognition and emotion, 25, 478-489.
Ekman, P., Friesen, W., Ellsworth, P. (1982). What are the relative contributions of facial behavior and contextual information to the judgment of emotion? In: Ekman, P., editor. Emotion in the human face. New York: Cambridge University Press. Lang, P. J., Bradley, M. M., & Cuthbert, B. N. (1990). Emotion, attention, and the startle
reflex. Psychological Review, 97, 377-395.
Mobbs, D., Weiskopf, N., Lau, H. C., Featherstone, E., Dolan, R. J., & Frith, C. D. (2006). The Kuleshov Effect: the influence of contextual framing on emotional attributions.
SCAN, I, 95-106.
Paulus, A., & Wentura, D. (2013). Threatening joy: Approach and avoidance reactions to emotions are influenced by the group membership of the expresser. Cognition &
Emotion. 1-22.
Phaf, R. H., Mohr, S. E., Rotteveel, M., & Wicherts, J. M. (2014). Approach, avoidance, and affect: a meta-analysis of approach-avoidance tendencies in manual reaction time tasks.
Frontiers in psychology, 5.
predispose for arm flexion and extension. American Psychological Association, 4, 156-172.
Solarz, A. K. (1960). Latency of instrumental responses as a function of compatibility with the meaning of eliciting verbal signs. Journal of Experimental Psychology, 59, 239-245.
Tottenham, N., Tanaka, J. W., Leon, A. C., McCarry, T., Nurse, M., Hare, T. A., ...
& Nelson, C. (2009). The NimStim set of facial expressions: judgments from untrained research participants. Psychiatry research, 168, 242-249.
Trope, Y. (1986). Identification and inferential processes in dispositional attribution.
